Fast personalized CT dose calculations with GPUMCD.

Area of Science:

• Medical Physics
• Radiological Sciences
• Computational Imaging

Background:

• Rising population radiation dose from Computed Tomography (CT) necessitates personalized dose estimation.
• Accurate patient-specific dosimetry is crucial for optimizing radiation safety in medical imaging.

Purpose of the Study:

• To demonstrate a GPU-enabled pipeline for patient-specific CT dosimetry.
• To combine automatic segmentation with GPU Monte Carlo Dose (GPUMCD) simulations.
• To generate dose-to-organ CT dosimetry reports from existing patient CT images.

Main Methods:

• A dynamic CT imaging process was simulated using GPUMCD and DICOM header information.
• A pipeline integrated GPUMCD with the TotalSegmentator (nnU-net) for automatic segmentation.
• Dose-to-organ dosimetry was performed on lung cancer screening CT images and reported in DICOM Structured Reports.

Main Results:

• GPUMCD dose calculations showed excellent agreement with phantom measurements (within 5.5%).
• The automated pipeline achieved an average runtime of 6 minutes and 6 seconds per CT study using one GPU.
• GPU simulation and segmentation accounted for 46% of the total pipeline runtime.

Conclusions:

• The pipeline enables rapid generation of patient-specific 3D dose distributions in CT scans.
• Fully automated, mass personalized dose-to-organ calculations are feasible.
• The DICOM-compliant output facilitates integration into clinical workflows and large-scale dosimetry studies.